Native support of mruby-cli in mruby build system.

[take-cheeze]

Just for a proof of concept I've rewritten mruby-bin-strip only with ruby script for mruby-cli feature.
For mrb_irep section modifications, MRubyVM (subset of RubyVM in CRuby) is implemented too.

Since mruby-cli isn't catching up with recent mruby releases, we should support this with mrbgem build system.
~~ This is not complete because I need to add minimal dependency mrb_open for better testings too. Symbol like mrb_open_cli must be generated. ~~
Now implemented: 34a3939

See #1676 for implementation detail.

[matz]

I am not sure the benefit of the proposed change. I try to read the diffs but they are huge.
Could you explain what we can expect from the change?

Besides that, this pull-request introduces 3 global variables. It's against our basic rule to avoid global variables.

[take-cheeze]

Ah, sorry these corresponds to those in CRuby:
https://github.com/ruby/ruby/blob/421a73f51a39ff6829f9715b423a928782ec8ec3/vm.c#L339-L341
And it should be mrb_state specific.

[take-cheeze]

Maybe this should be optional like VM hooks.

[take-cheeze]

@matz
This PR is mostly proof of concept so I'll split this into 3 PRs when approved:

• modification of build system(Rakefile, tasks/cli.rake, tasks/cli/main.c, toolchain script)
  • Forgot to add option to use mrb_open instead of minimal dependencies mrb_open_cli for command like mruby. Will implement it too.
• mrubyvm mrbgem which implements RubyVM class for mruby (related to: load bytecode from string (in ruby) #3955 )
• switchmruby-strip implementation to a mruby script

I want to know:

• modification to build system is OK.(1st PR of above)
• writing command line tools with mruby script is OK.

We need to know this feature works and all knows it's useful.
If OK I'll start splitting it to 3 PR.

[matz]

The proposed change sounds reasonable. It is OK to change the build system and writing command line tools in mruby.

But I don't want to add members to mrb_state unless they are absolutely necessary.